Design Article
Discrete audio amplifier basics - Part 1: Bipolar junction transistor circuits
John Linsley Hood
3/10/2010 2:00 PM EST
9.1 Junction Transistors
These are nearly always three-layer devices, fabricated by the multiple and simultaneous vapor phase diffusion and etching of small and intricate patterns on a large, thin slice of very high purity single crystal silicon. A few devices are still made in germanium, mainly for replacement purposes, and some VHF components are made in gallium arsenide, but these will not, in general, lie within the scope of this book.
The fabrication techniques may be based on the use of a completely undoped (intrinsic) slice of silicon, into which carefully controlled quantities of impurities are diffused through an appropriate mask pattern from both sides of the slice. These are described in the manufacturers' literature as double diffused, triple diffused, and so on.
In a later technique, evolved by the Fairchild Instrument Corporation, all the diffusions were made from one side of the slice. These devices were called planar and had, normally, a better HF response and more precisely controlled characteristics than, for example, equivalent double-diffused devices.
In a further, more recent, technique, also due to Fairchild, the silicon slice will have been made to grow a surface layer of uniformly doped silicon on the exposed side (which will usually form the base region of a transistor) and a single diffusion was then made into this doped layer to form the emitter junction. This technique was called epitaxial and led to transistors with superior characteristics, especially at HF. Since this is the least expensive BJT fabrication process, it will normally be used wherever it is practicable, and if no process is specified it may reasonably be supposed to be a planar-epitaxial type.
In contrast to a thermionic valve, which is a voltage-controlled device, the BJT is a current operated one. So while a change in the base voltage will result in a change in the collector current, this has a very nonlinear relationship to the applied base voltage. In comparison to this, the collector current changes with the input current to the base in a relatively linear manner. Unfortunately, this linear relationship between Ic and Ib tends to deteriorate at higher base current levels, as shown in Figure 9.1.
Figure 9.1: BJT nonlinearity.
This relationship between base and collector currents is called the current gain, and for AC operation is given the term hfe, and its nonlinearity is an obvious source of distortion when the device is used as an amplifier. Alternatively, one could regard this lack of linearity as a change of hFE (this term is used to define the DC or LF characteristics of the device) as the base current is changed. A further problem of a similar kind is the change in hfe as a function of signal frequency, as shown in Figure 9.2.
Figure 9.2: Decrease in hfe with frequency.
However, as a current amplifier (which generally implies operation from a high impedance signal source) the behavior of a BJT is vastly more linear than when used as a voltage amplifying stage, for which the input voltage/output current relationships are shown for an NPN silicon transistor as line 'a' in Figure 9.3. (I have included, as line 'b' , for reference, the comparable characteristics for a germanium junction transistor, although this would normally be a PNP device with a negative base voltage, and a negative collector voltage supply line.) By comparison with, say, a triode valve, whose anode current/grid voltage relationships are also shown as line 'c' in Figure 9.3, the BJT is a grossly nonlinear amplifying device, even if some input (positive in the case of an NPN device) DC bias voltage has been chosen so that the transistor operates on a part of the curve away from the nonconducting initial region.
Figure 9.3: Comparative characteristics of valve, germanium, and silicon based BJTs.
|
|
|
|
|
abraxalito
3/10/2010 9:05 PM EST
I thought that Doug Self as editor would have caught the myth being propagated here that the BJT is a 'current operated device'. He makes plain in his own writings that its not, rather its voltage operated with an exponential law between VBE and IC.
Sign in to Reply
d-jeff
3/11/2010 3:32 AM EST
The bipolar IS voltage controlled with an exponential law (even though when a VBE is applied, some IB takes place). In this article on bipolar, this fact does not seem obvious, I agree with previous comment.
Sign in to Reply
bcarso
3/11/2010 12:17 PM EST
Hood was a delightful character, but wasn't always entirely clear on solid-state physics. I recall at one point in his description of JFET operation he refers to majority carriers "tunneling" through the pinched-off channel...
Although there are situations where a voltage-controlled analysis can be better suited to a given device, Barrie Gilbert makes the point that really both voltage and current are always involved (see his way-best-of-book articles in Toumazou's (et al., eds) collection Analogue IC Design: the current-mode approach). Another very powerful and seldom-referenced approach, analysis in terms of charge as the variable, is presented in Ed Cherry's Amplifying Devices and Low-pass Amplifier Design, which does an integrated presentation of tubes and transistors.
An aside: Barrie G. also mentioned once some experimental heterojunction ADI bipolars that had a beta of order 100k.
Sign in to Reply
Guru of Grounding
3/11/2010 7:32 PM EST
Arguing about whether a bipolar transistor is voltage or current "operated" is like arguing about how many angels can dance on the head of a pin! Notice that the context of his statement included the word "linearity". Many transistors have very linear hFE over many decades of current while Ic vs Vbe is anything but linear. Linearity is THE prime consideration in audio after all. Circuit designers simply want to characterize a transistor as a 3-terminal device and, for linear applications, the quantum physics involved inside are of little interest. I have disagreements with Mr. Hood, but they're about other issues. I think his explanation is quite appropriate in its context.
Bill Whitlock
president/chief engineer
Jensen Transformers
www.jensen-transformers.com
Sign in to Reply
bcarso
3/16/2010 4:06 PM EDT
Agreed Bill---although I would suggest that the expression should be "very constant beta" rather than "linear" beta, as a function of collector current. "Linear" beta for example would be where beta was given by an expression like m*Ic + b.
Sign in to Reply
c_abraham
6/12/2010 7:07 PM EDT
Then why does every semiconductor OEM call the bjt a current controlled device? Do you know more then them? How many semicond physics courses have you taken? Nothing personal, but if what you say is true, every OEM would concur, & every university semicond course would teach the same.
At the "black box" viewpoint, a bjt is classified as current controlled. For a deeper look involving internal charge distribution, depletion regions, etc., the charge control model is used, for the bjt & FET as well. Ultimately, at the atomic level, only quantum mechanics can describe what is happening. This is the universal model. Shockley stated this in the early 1950's.
Again, at the external viewpoint, current control is a good 1st order approximation. More in depth study produced the charge control model. Ultimately the device is best described by QM. Every OEM cannot be wrong. I took 1 undergrad course in semicond phy from the phy dept. Then I took 4 couses in semicond phyfrom the EE dept. in my MS & Ph.D. studies at 2 different universities. The bjt & FET are described as I just stated.
A bjt cannot be directly driven from a low impedance voltage source. Thermal runaway would take place. We always "current drive" a bjt. If the input signal source is constant voltage type, then a resistor must be used to drive the bjt. This is why bjt is "current driven". The Vbe is of course, absolutely needed, as are Ib & Ie. All 3 are needed for transisto action to occur. Without Ib, Vbe, & Ie, there can be no Ic. They are equally important.
Claude
Sign in to Reply
Kevin Aylward
6/2/2013 7:02 AM EDT
1 Semiconductor companies possibly just repeat the old wife’s tales that they know about. Possible its because good analog designers are hard to come by.
2 University semiconductor courses do indeed teach that the transistor is a voltage controlled device.
3 The Gummel-Poon charge controlled model expressly describes the transistor as a voltage controlled device. This is explained here:
http://www.kevinaylward.co.uk/ee/voltagecontrolledbipolar/voltagecontrolledbipolar.html
4 Driving a bipolar from a current source for biasing results in unpredictable biasing. This is explained here:
http://www.kevinaylward.co.uk/ee/bipolardesign3/bipolardesign3.html
5 Bipolors are easily driven from low impedance sources, e.g. Emitter resisters can be used when driving bipolars from large, voltage driven sources. Additionally, adding large resistors to the base circuit results in more noise, and lower bandwidth.
6 I could go on, but the original comments above are, essentially, incorrect.
Sign in to Reply
pcsalex
7/15/2011 9:47 AM EDT
by the way on figure 7B that is not a rearranged cascaded, but a differential amplifier!
Sign in to Reply